Committed my original patch in r205398. Happy to bikeshed further, revert, etc, if anyone has better ideas/disagreements/etc.
On Wed, Mar 12, 2014 at 5:25 PM, Arthur O'Dwyer <[email protected]> wrote: > Ah, if only C++ hadn't already soaked up every possible delimiter-pair > ((), [], {}, <>, [[]], and in various extensions <<<>>>, ({}), (()), > ...). No matter what you pick, it'll collide with *some* C++ syntax. > > Insert obligatory Unicode proposal here. ;) > > FWIW (not much), I like David's original proposal of ()s instead of > <>s, because IF the goal is to avoid user-confusion, and IF we > postulate that the user-confusing-est error messages tend to come from > deeply nested templates, then <>s are worse than ()s. > > -Arthur > > > On Wed, Mar 12, 2014 at 4:58 PM, David Majnemer > <[email protected]> wrote: >> FYI, GCC seems to use curly braces for some diagnostics: >> >> given: >> namespace { int x; } >> namespace { int x; } >> >> GCC outputs: >> error: redefinition of 'int {anonymous}::x' >> >> Personally, I like the curly braces more than the parens. >> >> Just my two cents on the issue. >> >> On Wed Mar 12 2014 at 4:51:39 PM, Richard Smith <[email protected]> >> wrote: >>> >>> On Tue, Mar 11, 2014 at 12:08 AM, David Blaikie <[email protected]> >>> wrote: >>>> >>>> Partly out of consistency with GCC's anonymous namespace rendering >>>> ("(anonymous namespace)") and just a desire not to have angle brackets >>>> be confused with a template argument list, this patch changes all our >>>> anonymous things to use parens. >>>> >>>> Not entirely consistent with GCC (GCC still uses angles for >>>> "<anonymous struct>" and other cases), but Chandler and Richard seemed >>>> to prefer consistency between the different kinds of names, rather >>>> than consistency with GCC here. >>>> >>>> I haven't added an actual debug info test, though that was my initial >>>> motivation - I can add one if someone thinks that's valuable. >>>> >>>> Any other thoughts? >>> >>> >>> Looking through the testcase updates, my gut reaction is that I find the >>> ()s uglier than the <>s (maybe because it's more obvious that the <>s are >>> not real language syntax?)... but I'm sure I'll get over it :) >>> _______________________________________________ >>> cfe-commits mailing list >>> [email protected] >>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits >> >> >> _______________________________________________ >> cfe-commits mailing list >> [email protected] >> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits >> _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
